OpenAI builds its own AI chip 'Jalapeño' in 270 days: when AI companies stop relying solely on Nvidia

OpenAI unveiled its first custom-designed AI chip, named 'Jalapeño,' built in partnership with Broadcom. The chip, developed in just 270 days, is designed specifically for AI inference workloads and marks a strategic shift away from sole reliance on Nvidia GPUs.

What happened

OpenAI announced on June 24, 2026 that it had successfully developed its first custom AI chip, codenamed "Jalapeño," in collaboration with Broadcom. The chip went from design to tape-out in 270 days — an unusually fast timeline for semiconductor development — with OpenAI's own AI models assisting in the design process. Mass production is expected to begin in 2027, with Samsung supplying HBM4 memory under an exclusive deal.

Knowledge point: why AI companies want their own chips

The AI industry's explosive growth has created an unprecedented demand for computing power. Nvidia's GPUs have been the default choice because their parallel processing architecture is well-suited for training and running large AI models. However, relying on a single supplier creates bottlenecks: allocation constraints, pricing power, and design trade-offs that favor general-purpose workloads over specific AI inference needs. Custom chips allow companies to optimize for their exact requirements — lower power consumption, lower latency for inference, and better cost efficiency at scale.

The vertical integration trend

OpenAI joins a growing list of tech giants — including Google (TPU), Amazon (Trainium/Inferentia), Microsoft, and Meta — that have invested in custom silicon. The knowledge lesson: in an industry where compute is the new oil, controlling your chip supply is not just a cost-saving measure but a strategic necessity. The shift from buying GPUs to building custom chips represents a fundamental restructuring of the AI supply chain.